How to test multiple explanations against evidence instead of letting confirmation bias decorate your favorite story.

Analysis of Competing Hypotheses, or ACH, is a disciplined method for comparing multiple explanations against available evidence. It is especially useful when people are arguing about why something happened, what is really going on, or which explanation best fits the facts.

The ordinary human method is to pick a favorite explanation, collect confirming evidence, ignore awkward evidence, and call this “being reasonable.” ACH does something better. It asks which hypothesis has the fewest serious conflicts with the evidence. This is not glamorous, but neither is cleaning the kitchen. Both prevent disease.

 

 

Quick navigation

 

Best used for

    • Investigating disputed causes.
    • Testing competing interpretations.
    • Reducing confirmation bias.
    • Choosing between explanations in politics, work, research, relationships, or strategy.
    • Identifying what evidence would most reduce uncertainty.

 

5-minute version

Use this when the problem is pressing and you need the fastest responsible version of the method. Not perfect, but better than sprinting into a decision while waving a flaming assumption.

    1. Write the question you are trying to answer.
    2. List at least three plausible explanations.
    3. List the strongest evidence you have.
    4. Ask which evidence conflicts with each explanation.
    5. Temporarily favor the explanation with the fewest serious conflicts, not the one you like most.

 

 

30-minute careful version

Use this when the issue matters enough to deserve a slower look. Thirty minutes of structured thinking can prevent thirty months of cleanup, which is apparently a bargain humans keep trying to avoid.

    1. Define the analytic question clearly.
    2. List all plausible hypotheses, including unpopular or uncomfortable ones.
    3. List evidence and assumptions separately.
    4. Create a simple matrix: evidence down the left, hypotheses across the top.
    5. Mark whether each piece of evidence is consistent, inconsistent, or not diagnostic for each hypothesis.
    6. Focus on disconfirming evidence because it is more useful than evidence that fits everything.
    7. Identify what new evidence would most reduce uncertainty.
    8. State a provisional conclusion with confidence level and remaining unknowns.

 

 

 

 

Vignette: The mysterious membership decline

An organization sees membership decline. One person blames the website. Another blames the economy. Another blames messaging. Another quietly suspects the organization has become less useful, which is rude of reality but possible.

ACH forces them to compare each explanation against actual evidence: traffic data, renewal surveys, donor comments, program attendance, competitor growth, email open rates, and interviews. The website is ugly, yes, but the decline started before the redesign. The strongest conflict is with the “website only” hypothesis. The better explanation is reduced perceived value plus weak follow-up. The website is not innocent, but it is not the main criminal.

 

Practice: apply this to one of your three current problems

Write down your three most important current problems. Pick one. Then apply the prompts below. Do not merely admire the tool from a safe distance like a museum visitor staring at a fire extinguisher.

    1. Pick one disputed explanation in your life or work.
    2. List at least three possible explanations.
    3. Write five pieces of relevant evidence.
    4. For each explanation, ask: does this evidence support it, conflict with it, or fail to distinguish it?
    5. Write the current best explanation and one thing that could change your mind.

 

 

Common mistakes

    • Using ACH to prove your favorite hypothesis.
    • Listing only two explanations when more are plausible.
    • Confusing evidence with interpretation.
    • Ignoring evidence that conflicts with the explanation you prefer.
    • Treating weak evidence as decisive because it is emotionally satisfying.

 

AI Prompt Support Module

Use AI as a thinking partner, not as a priest, judge, or magical vending machine for certainty. First, write your own answer. Then ask AI to challenge, improve, and stress-test it.

Build an ACH matrix

I am trying to answer this question: [question]. Generate at least five plausible hypotheses and help me build an ACH matrix using the evidence I provide: [evidence]. Focus especially on which evidence conflicts with each hypothesis.

Find missing explanations

Here is my favored explanation: [explanation]. Generate alternative explanations I may be overlooking. Include at least one mundane explanation, one systemic explanation, one incentive-based explanation, and one uncomfortable explanation.

Stress-test evidence

Here is a list of evidence: [list]. Tell me which pieces are diagnostic, which fit too many hypotheses to be useful, which may be weak or biased, and what new evidence would most reduce uncertainty.

 

FAQ

Is ACH only for intelligence analysts?

No. It was developed in intelligence analysis, but it is useful anywhere several explanations fit some of the evidence.

Why focus on disconfirming evidence?

Because confirming evidence often fits many hypotheses. Evidence that seriously conflicts with one explanation helps narrow the field.

Can ACH prove the truth?

Usually no. It improves comparative judgment. It helps you state which explanation currently fits best and what remains uncertain.

 

Glossary

    • Hypothesis: A possible explanation for what is happening or why something happened.
    • Diagnostic evidence: Evidence that helps distinguish between competing hypotheses.
    • Disconfirming evidence: Evidence that weakens or conflicts with a hypothesis.
    • Confidence level: A stated degree of certainty, such as low, moderate, or high, based on evidence quality and uncertainty.

 

 

References and bibliography

These sources are included so readers can go deeper, examine the intellectual foundations, and avoid treating this guide as if it descended from the clouds on a glowing clipboard.

    1. Richards J. Heuer Jr., Psychology of Intelligence Analysis, CIA Center for the Study of Intelligence. CIA PDF.
    2. Amos Tversky and Daniel Kahneman, “Judgment under Uncertainty: Heuristics and Biases,” Science, 1974. PubMed record.

 

Next: Forecasting and Calibration

The next page turns from explaining what may be happening now to estimating what may happen next. Forecasting is not fortune-telling with better stationery. It is disciplined uncertainty management.

You will learn how to make clearer forecasts, use base rates, update as evidence changes, and score your predictions so your confidence learns manners.